Novel Deep Learning Reconstruction to Augment Contrast Enhancement: Initial Evaluation

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Single-energy CT (SECT) and dual-energy CT (DECT) are crucial for visualizing abdominal pathologies, particularly colorectal adenocarcinoma and liver metastases.
• Improving contrast enhancement in SECT is essential for accurate diagnosis and treatment planning.
• Deep learning (DL) offers potential for enhancing CT image quality and diagnostic performance.

Purpose of the Study:

• To evaluate the image quality of a novel deep learning (DL) reconstruction for SECT.
• To compare the contrast enhancement achieved by DL-reconstructed SECT with standard SECT and DECT.
• To assess improvements in artifacts, noise texture, and resolution using the DL reconstruction.

Main Methods:

• Retrospective analysis of raw data from a prospective study involving patients with colorectal adenocarcinoma and liver metastases.
• Acquisition of 120 kVp SECT and 50 keV DECT abdominal scans in the portal venous phase.
• Application of a novel DL algorithm for SECT reconstruction and independent assessment by two readers.

Main Results:

• The DL reconstruction demonstrated significantly higher Hounsfield Units (HUs) in liver, pancreas, spleen, psoas muscle, and aorta compared to standard 120 kVp SECT.
• Hounsfield Units with DL reconstruction were significantly lower than with 50 keV DECT.
• Readers rated the DL reconstruction as having superior contrast enhancement, improved artifact reduction, better noise texture, and enhanced resolution compared to standard 120 kVp SECT.

Conclusions:

• The novel DL reconstruction for SECT provides superior contrast enhancement compared to standard 120 kVp SECT.
• The DL method approaches the contrast enhancement levels of 50 keV DECT.
• The DL reconstruction offers significant improvements in perceived artifacts, noise texture, and resolution.